Technology Report Received: May 20, 2014 Accepted after revision: October 26, 2014 Published online: February 18, 2015
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Using the Accelerometers Integrated in Smartphones to Evaluate Essential Tremor Suhan Senova a Damien Querlioz c Claire Thiriez b Pierre Jedynak a Béchir Jarraya a Stéphane Palfi a
a
Neurosurgery Department and b Neurology Department, Assistance Publique-Hôpitaux de Paris (APHP), Groupe Henri-Mondor Albert-Chenevier, Université Paris 12 UPEC, Faculté de Médecine, Créteil, and c Institut d’Electronique Fondamentale, Université Paris-Sud, CNRS, Orsay, France
Abstract Background/Aims: Evaluation of tremor constitutes a crucial step from the diagnosis to the initial treatment and follow-up of patients with essential tremor. The severity of tremor can be evaluated using clinical rating scales, accelerometry, or electrophysiology. Clinical scores are subjectively given, may be affected by intra- and interevaluator variations due to different experience, delays between consultations, and subtle changes in tremor severity. Existing medical devices are not routinely used: they are expensive, timeconsuming, not easily accessible. We aimed at showing that a smartphone application using the accelerometers embedded in smartphones is effective for quantifying the tremor of patients presenting with essential tremor. Methods: We developed a free iPhone/iPod application, Itremor, and evaluated different parameters on 8 patients receiving deep brain stimulation of the ventral intermediate nucleus of the thalamus: average and maximum accelerations, time above 1 g of acceleration, peak frequency, typical magnitude of tremor, for postural and action tremors, on and off stimulation. Results: We demonstrated good correlations between the parameters measured with Itremor and clinical score in all conditions. Itremor evaluation enabled higher discriminatory
© 2015 S. Karger AG, Basel 1011–6125/15/0932–0094$39.50/0 E-Mail [email protected] www.karger.com/sfn
power and degree of reproducibility than clinical scores. Conclusion: Itremor can be used for routine objective evaluation of essential tremor, and may facilitate adjustment of the treatment. © 2015 S. Karger AG, Basel
Introduction
Tremor constitutes a major motor symptom of several neurological diseases such as Parkinson’s disease and essential tremor. The severity of tremor can be evaluated using clinical rating scales, such as the Fahn-TolosaMarin score (Fahn score) [1]. Tremor can also be considered as a physical phenomenon, characterized through actigraphic, accelerometric, electromyographic, gyroscopic, or digitizing tablet recordings [2–11]. Precise evaluation of tremor is crucial for the diagnosis of neurological diseases, patient follow-up and the evaluation of treatment efficacy [12, 13]. The current clinical rating scales suffer from intra- and interrater variability that depends on several factors including rater experience and the delays between consultations. Furthermore, rating scales may not be sensitive enough to detect small variations of tremor [14–16]. Recordings made with current medical devices, while more objective, are not an optimal way to assess tremor either in clinical routine or for regular follow-up. Indeed, these devices are expensive, Suhan Senova and Pr. Stéphane Palfi Neurosurgery Department, Assistance Publique-Hôpitaux de Paris (APHP) Groupe Henri-Mondor Albert-Chenevier, Université Paris 12 UPEC Faculté de Médecine, FR–94010 Créteil (France) E-Mail stephane.palfi @ hmn.aphp.fr
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
Key Words Tremor · Essential tremor · Movement disorders · Accelerometry · Smartphone
from the three spectra. This analysis is only valid if the motion recorded is oscillatory enough. The application thus considered the spectrum analysis as valid only if the peaks were higher than an empirical threshold (equal to 15% of the spectral peak of patients with severe essential tremor scored 4/4 with Fahn score). When this validation criterion was met, the peak frequency parameter in Itremor was given by the average of its values in the three directions. Finally, the typical magnitude was computed in centimeters as mean acceleration divided by squared frequency and multiplied by a conversion factor taking into account scale changes and integration operation. The typical magnitude was meant as an estimation of the peak-to-peak movement amplitude of the hand in the tremor. Control experiments showed that this evaluation was effective for oscillatory movements. When the validation criterion was not met, the tremor magnitude was assigned the value 0.3 mm, which is the minimal distance of motion the human eye can see at a distance of 1 m. The five parameters, as well as the power spectra, were computed within this application Itremor and were accessible directly on the iPhone/iPod touch screen.
Apparatus We used an Apple iPod touch device. It is equipped with a three-axis accelerometer manufactured by ST Microelectronics as a standard feature. Our application performed recordings during 9-second periods of time with a sampling rate of 60 Hz. Within the application, it is possible to activate an ‘action tremor analysis’ mode, for which accelerometric data are high-pass filtered (one pole, 2.5-Hz cutoff) for the component of the signal due to the voluntary motion to be removed. When this mode was not activated, a 1-Hz cutoff was used to filter gravity. From the raw data recorded by the three-dimensional accelerometer, the application computed five parameters: the mean acceleration, the maximum acceleration, the time above 1 g, the peak frequency, and the typical magnitude of the tremor. The mean and maximum accelerations were computed directly from the filtered accelerometric data. The parameter time above 1 g is the time during each recording for which values of acceleration measured are greater than 9.8 m/s2, a standard accelerometric figure. The last two parameters were calculated by performing a spectral analysis. The spectrum density of the acceleration along the three x, y and z directions was computed with fast Fourier transform using Welch’s method with three overlapping segments and a Hamming window function. The highest peak was extracted
Clinical Assessment Eight patients, 3 women and 5 men, with an average age of 61 ± 13.5 years, were evaluated during a routine follow-up outpatient visit. This group of patients was homogeneously made of patients presenting with essential tremor. They all signed an informed consent for tremor assessment by Itremor and a videotaping for a blinded evaluation using the Fahn score. We received the agreement of the Ethics Committee of Henri Mondor University Hospital/Paris 12 Créteil, France. These patients suffered from essential tremor and were treated by uni- or bilateral Vim high-frequency electrical stimulation (table 1). Electrode implantation was previously described [19]. The iPod touch was attached to a dedicated armband tightly fixed around the wrist displaying the most severe tremor in the offstimulation condition (fig. 1). This armband is a commercial solution proposed by Apple in order to hold strongly the iPod while performing sport for instance. For objective evaluation of Itremor, during the consultation tremor was evaluated in on- and off-stimulation conditions while the patient was videotaped for subsequent blinded clinical score rating: thus, each patient was his own control. We evaluated postural tremor in the outstretched arms condition 3 times in a row, getting a triplet of values, and then, as a complement, in the swordsman condition also 3 times in a row. The swordsman condition consists in holding arms horizontally, with the elbows in semiflexion and supination, and with the fingers of each hand facing each other (fig. 1a) [20]: both distal and proximal postural tremors are involved. Results from outstretched arms and swordsman conditions were gathered under the name of postural tremor. Hand action tremor was evaluated 3 times in a row by asking the patients to go, pick up and give back a pen several times to the clinician in front of him. Between each evaluation, the patients could rest 1 min. Between the on- and off-stimulation series, the patients could rest 3 min. At the end, for subjective evaluation of Itremor, patients were asked the following questions: ‘Was the evaluation with Itremor tolerable?’ ‘Do you understand the use of Itremor?’ ‘Would you accept that your regular follow-up includes an evaluation with Itremor?’
Evaluation of Essential Tremor with Smartphones
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Methods
95
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
time-consuming and not easily accessible. The procedure may be poorly tolerated by the patient if performed on a routine basis, and the clinicians are required to learn elaborate connectors and software. As a consequence, specific recordings of tremor are now rarely resorted to, whereas they could be useful for both routine clinical follow-up and clinical research. For example, patients who are treated with deep brain stimulation of the ventral intermediate nucleus of the thalamus (Vim) for invalidating tremor need several consultations with a movement disorder specialist for adjusting to optimal stimulation parameters. Thus, there is a critical need for validating a medical device capable of evaluating tremor features objectively in clinical routine, being inexpensive, easily accessible, easy to use, and well tolerated by the patient. To address this clinical challenge, we developed a free iPhone/iPod application called Itremor that uses built-in accelerometers embedded in smartphones/MP3 players. We aimed at bringing the proof of concept that a smartphone application is effective for quantifying the tremor of patients diagnosed with essential tremor. We evaluated the clinical value of Itremor in such patients treated with electrical stimulation of the Vim. This is in line with the current idea of resorting to smartphones for medical and research purposes [17, 18].
b
Fig. 1. a Patient equipped with an armband containing the iPod and tightly fixed to her wrist. She is holding the swordsman position. b, c Printscreens of the iPod when the iPod application Itremor was used. b The tremor was recorded in the off-stimulation
c
Color version available online
a
condition and was scored 4/4 blindly by a neurologist. c The tremor was recorded in the on-stimulation condition and was scored 0/4 blindly by a neurologist. Each curve (red, green or blue) refers to one direction of space (colors refer to the online version only).
Table 1. Characteristics of the population of the study
Patient
Gender
Age, years
Date of start of stimulation
Recorded side
Parameters of monopolar stimulation: frequency (Hz)/pulse width (μs)/amplitude (V)/ impedance (Ω)/electrode site
1
female
60
2009
R
L: 130/60/2.5/491/1– R: 130/60/2.8/403/4–
2
male
58
1999
L
R: 185/90/3.2/127/4–
3
female
78
2008
L
L: 130/90/3.4/1,360/0– R: 130/90/2.5/491/4–
4
male
82
2006
R
L: 180/90/2.7/532/0– R: 180/90/2.3/450/5–
5
male
48
2010
L
L: 185/90/2.6/1,035/0– R: 185/90/2.3/913/4–
6
female
42
2010
R
L: 130/90/0.5/4,005/0– R: 130/90/1.8/711/4–
7
male
58
2006
L
L: 160/90/2.6/1,035/0– R: 160/90/2.8/267/4–
8
male
61
2010
L
L: 240/90/1.9/369/0– R: 240/90/1.9/751/4–
96
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
L = Left; R = right.
Color version available online
Itremor parameters
–4
n = 96 0
Action
log (Itremor parameter)
0
1 2 3 Clinical score (/4)
4
R = 0.83 p < 0.0001
–1
–2 n = 48 0
1
2
3
Clinical score (/4)
0 –1 –2 –3
n = 96 0
b
4
0
–1 n = 48 1
2
3
Clinical score (/4)
Fig. 2. Scatter plots (dots) and linear regression fit lines showing sig-
nificant correlation (R) between the Fahn clinical score and the natural logarithm of various Itremor parameters: mean acceleration (expressed in g), maximum acceleration (expressed in g) and tremor magnitude (expressed in cm), in two different conditions (postural,
Itremor data were stored at the end of each evaluation. Fahn scores were obtained blindly after randomization of the videos by a senior neurology specialist of movement disorders (P.J.). All videos were rated during the same session. Data Analysis Methods We first evaluated the clinical relevance of the parameters displayed by Itremor. We calculated the correlation between Fahn scores and the natural logarithm of mean acceleration, the natural logarithm of maximum acceleration, and the natural logarithm of typical magnitude of the tremor in the postural condition and in the action condition [21], pooling on- and off-stimulation condition measures from all the patients. Then, within each subgroup of measurements corresponding to each of the five grades of the Fahn scale, we calculated the percentage of measurements for which time above 1 g was superior to 1% of the time of recording, and for which a spectral analysis had been computed by Itremor. We resorted to Matlab software (MathWorks, Natick, Mass., USA) for this purpose. Finally, we compared the performance of the Fahn scores established by a movement disorder specialist expert in tremor and the Itremor measures as far as discriminatory power and degree of reproducibility (DR) of the measures are concerned. To this pur-
Evaluation of Essential Tremor with Smartphones
4
R = 0.71 p < 0.0001
1
0
e
1 2 3 Clinical score (/4)
R = 0.86 p < 0.0001
2 1 0 –1
n = 96 0
c
f
4
R = 0.83 p < 0.0001
1 0 –1
n = 48 0
4
1 2 3 Clinical score (/4)
1
2
3
4
Clinical score (/4)
action). For each linear regression, trials from all the patients and in on- and off-stimulation conditions have been pooled (n = 96 trials for the postural condition, n = 48 trials for the action condition). Correlation between Itremor parameters and clinical score was better for postural tremors (a–c) than for action tremor (d–f).
pose, for each condition and for each method of tremor evaluation, we computed areas under receiver operating characteristic (ROC) curves as a measure of discriminatory power [22] and performed statistical comparison between areas under paired ROC curves, resorting to MedCalc software (Mariakerke, Belgium) [23]. We defined DR as the ratio between the mean of the measure of all the trials and the mean of the triplet of the standard deviation of the measure. For a given condition, variation of DR was the ratio of the difference between the DR for a given parameter measured by Itremor and the DR of the Fahn score, divided by the DR of the Fahn score.
Results
In the postural condition, when pooling on- and offstimulation data from all the patients, a significant correlation was found (fig. 2) between Fahn scores and the natural logarithm of mean acceleration (r = 0.92, p < 10–4, n = 96), Fahn scores and the natural logarithm of maximum acceleration (r = 0.89, p < 10–4, n = 96), Fahn scores Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
97
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
–3
R = 0.89 p < 0.0001
log (Itremor parameter)
–2
a
d
log (Itremor parameter)
–1
log (Itremor parameter)
Postural
log (Itremor parameter)
0
Tremor magnitude
Maximum acceleration
1
R = 0.92 p < 0.0001
log (Itremor parameter)
Mean acceleration
1
Action
80 Sensitivity
40
0
0 0
60
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
0
100 – Specificity
and the natural logarithm of tremor magnitude (r = 0.86, p < 10–4, n = 96). Similarly, in the action condition, a significant correlation was found (fig. 2) between Fahn scores and the natural logarithm of mean acceleration (r = 0.83, p < 10–4, n = 48), Fahn scores and the natural logarithm of maximum acceleration (r = 0.71, p < 10–4, n = 48), Fahn scores and the natural logarithm of tremor magnitude (r = 0.83, p < 10–4, n = 48). For all the clinical situations rated 0, 1 or 2/4 on the Fahn tremor rating scale, the time above 1 g was never more than 1% of the recording time. This parameter was relevant to assess tremors rated 3 or 4/4 (table 2). When a tremor was scored 0 or 1/4 on the Fahn scale, it never passed the validation criterion which allows spectral analysis to be computed (table 2). Discriminatory powers between on- and off-stimulation condition recordings were significantly higher (p < 0.05) for the parameters mean acceleration and maximum acceleration from Itremor than for the Fahn score in the postural condition (fig. 3). We observed that the DR of the mean acceleration and the tremor magnitude were improved in all conditions compared with the DR of the Fahn score (table 3). The DR of the maximum acceleration was lowered compared with the DR of the Fahn score for the action condition, and was improved for the postural condition. A hundred percent of patients found that the evaluation with Itremor was tolerable, understood the use of Itremor and accepted a regular evaluation with Itremor during follow-up. 98
40
60 100 – Specificity
Clinical score log (mean acceleration)
log (maximum acceleration) log (tremor magnitude)
Table 2. Relevance of the Itremor parameter time above 1 g and the spectral analysis for each grade of the Fahn scale in two conditions (postural, action) Clinical score
0 (n = 26 trials) 1 (n = 56 trials) 2 (n = 32 trials) 3 (n = 15 trials) 4 (n = 15 trials)
Trials for which time above Trials for which spectral 1 g >1% of recorded time, % analysis was performed, % postural
action
postural
action
0 0 0 10 100
0 0 0 60 100
0 0 78 100 100
0 0 46 100 100
Discussion
All the parameters given by Itremor were clinically correlated for both the postural and action components of tremor in patients suffering from essential tremor. The use of Itremor for tremor assessment was well tolerated and well accepted by patients. To define the parameters to be determined by Itremor, we were guided by the functional impact of tremor, looking at patients drinking a glass of water. The difficulties encountered are mostly due to the acceleration to which the water is submitted inside the glass without spilling over and to the magnitude of the oscillations of the glass when approaching the mouth. Thus, we calculated parameters directly related to the acceleration such as the mean acceleration, the maximum acceleration and the time above 1 g, and also parameters coming from the Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
tremor assessment, for postural and action tremor separately. Curves for which the area under the ROC curve is statistically (p < 0.05) different from the area under the ROC curve for clinical score rating are indicated with an asterisk. Discriminatory powers between on- and off-stimulation condition recordings were significantly higher (p < 0.05) for log (mean acceleration) and log (maximum acceleration) from Itremor than for the Fahn score in the postural condition.
Sensitivity
80
Fig. 3. ROC curves for each modality of
Color version available online
Postural
spectral analysis such as the peak frequency and the magnitude of the tremor. The armband to which the iPod was attached was fixed around the wrist, but could also be attached around the hand or the elbow depending on clinical specificities of the patient’s tremor. Mean and maximum accelerations are parameters that are simple to compute and to understand. But maximum acceleration may be biased by short periods of violent accelerations or by maximum acceleration exceeding the capacity of the accelerometer of the iPod touch. The parameter time above 1 g should enable us to better characterize and follow up patients with violent tremors; table 2 shows that this parameter is specifically relevant for tremors rated 3 or 4 on the Fahn scale. In clinical practice, treatment efficacy measured on a violent tremor may be better estimated by the reduction of periods of high acceleration. Three other parameters were computed after spectral analysis. For this kind of analysis to be physically valid, it has to be performed only in clinical situations where the motion of the wrist could be estimated as oscillatory, even if only transiently, by a clinician. From this idea, we elaborated a criterion on the value of the peak of the spectral analysis for the spectral analysis to be regarded as valid, then peak frequency and tremor magnitude to be calculated and displayed on the screen of the iPod. The validity of this criterion was confirmed by the results in table 2: no motion rated 0 or 1 on the Fahn scale justified spectral analysis, all motions rated 3 and 4 justified spectral analysis. In this study, we quantitatively compared the performance of Itremor with a validated clinical scale (Fahn score). We looked for a logarithmic correlation between mean acceleration, maximum acceleration, tremor magnitude calculated by Itremor and clinical scores, as observed in previously published studies with other tools of
tremor assessment [21, 24–28]. For all these parameters and in all the conditions, a significant correlation with the Fahn score was found. Testing patients in on- and offstimulation conditions enabled us to evaluate the correlation between Itremor parameters and clinical score, for tremors clinically scored from 0 to 4, as detailed in table 2. In the next study, we will evaluate whether using Itremor for the adjustment of deep brain stimulation parameters, more subtle than on- versus off-stimulation parameter adjustments, enables better outcomes of Vim deep brain stimulation. The study also demonstrates that a medical device with an accelerometer and appropriate data processing offers higher discriminatory power for postural tremor. We describe improved evaluation reproducibility of the tremor of a given patient by a given evaluator performing multiple evaluations at a given time. These findings were imaginable due to the finer graduations of an accelerometric data scale with a sensitivity of 20 mg of acceleration compared with a clinical scale with only five levels. Characterizing action tremor required not spoiling the analysis with the voluntary component of movement. Fortunately, this component was typically in a frequency range inferior to 2.5 Hz for the task we asked the patients to perform: so signal was high-pass filtered before the parameters were computed. The correlation coefficients between the various parameters from Itremor and the clinical scores were thus good even if slightly lower for action than for postural tremor: the frequential impoverishment of the signal after high-pass filtering might be one of the causes. However, the difficulty to score that type of tremor clinically may be another cause: during the movement, tremor may vary, and it may also be difficult for both the patient and the clinician to reproducibly perform the task. As far as discriminatory power is concerned, a superiority of Itremor parameters over clinical score was not found for action tremor: not only the lower correlation between Itremor parameters and clinical score for action tremor may account for it, but also comparing two sets of stimulation parameters and their clinical effect with less difference than on- versus off-stimulation parameters may have helped revealing a higher discriminatory power for Itremor parameters than for clinical score parameters when evaluating action tremor. Other groups tried to develop similar iPhone applications to quantify tremor [29, 30]. However, in contrast to the Itremor application, these applications were not yet validated clinically. In addition, our application included modalities for assessing not only postural trem-
Evaluation of Essential Tremor with Smartphones
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
celeration, maximum acceleration and tremor magnitude, relative to clinical Fahn score in two conditions (postural, action) Condition
Itremor parameter
Variation of DR, %
Postural (n = 96 trials)
mean acceleration maximum acceleration tremor magnitude
61.7 16.8 29.2
Action (n = 48 trials)
mean acceleration maximum acceleration tremor magnitude
61.6 –429.1 2.5
Data from all the patients in on- and off-stimulation conditions have been pooled.
99
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
Table 3. Variation of DR for three Itremor parameters: mean ac-
or as in the work of LeMoyne et al. [29] and Kostikis et al. [30], but also action tremors. Moreover, Itremor has the advantage of being completely autonomous and the whole signal analysis is performed and displayed on the iPhone/iPod directly, without requiring any data transfer to an external computer by e-mail [29], or being Web-based and running on a distant computer [30]. As a result, Itremor is easily usable in the context of outpatient clinics or hospitalization by clinicians and nurses. The iPhone application called ‘iSeismo’ was also used to quantify tremor [31]. However, this application was originally developed for earthquake detection and not for clinical use, whereas Itremor’s parameters aimed at quantifying specific clinically relevant motions, with algorithms adjusted to take into account clinical orders of magnitude. The first study showing a clinical validation of a smartphone application was published in 2013 [32]. In this study, the authors set up their clinical rating scale and did not refer to an internationally recognized clinical tremor rating scale. In addition, the author recognized their inability to detect short moments of intense and incapacitating tremor which were easily quantified by Itremor using the parameter called ‘time above 1 g’. Several other iPhone applications which were not clinically validated nor described in any scientific paper can be found on the Apple Store [Itrem (2011), ParkinsonMeter (2012), LiftPulse (2013) and StudyMyTremor (2014)]. Itremor does not compete with electromyography for precise diagnosis. They rather complement each other: Itremor may facilitate the routine objective eval-
uation of essential tremor and facilitate adjustment of the treatment. Itremor is a free application, distributed on request according to the Ad Hoc process defined by Apple. Smartphones are widely available and used by medical doctors and nurses who could have their own smartphone equipped with Itremor, ready to be used extensively and easily in the outpatient clinic and the inpatient department and even during clinical trials. Itremor is easy to use, and a detailed processing of data is done and displayed directly on the screen of the iPhone without the need for expert computer skills or signal processing. Data are also easy to save and export to a computer with secured transfer methods. Itremor highlights the potential impact of smartphones for modern, inexpensive, and efficient healthcare.
Conclusion
Evaluating essential tremor with the iPhone application Itremor was validated. It may improve tremor evaluation, follow-up, and treatment adjustment and facilitate clinical research. Itremor and electromyography complement each other and do not compete. Acknowledgment We thank the Association pour la Recherche sur la Stimulation Cérébrale (ARSC), Dr. Liza Leventhal, Dr. Olivier Bill and Arthur Silve for reviewing the manuscript.
1 Fahn S, Tolosa E, Marin C: Clinical rating scale for tremor; in Jankovic J, Tolosa E (eds): Parkinson’s Disease and Movement Disorders. Baltimore, Williams & Wilkins, 1993, pp 225–234. 2 Gerr F, Letz R, Green RC, et al: Relationships between quantitative measures and neurologist’s clinical rating of tremor and standing steadiness in two epidemiological studies. Neurotoxicology 2000;21:753–760. 3 Hoff JI, Wagemans EA, van Hilten BJ: Ambulatory objective assessment of tremor in Parkinson’s disease. Clin Neuropharmacol 2001; 24:280–283. 4 Van Someren EJ: Actigraphic monitoring of movement and rest activity rhythms in aging, Alzheimer’s disease, and Parkinson’s disease. IEEE Trans Rehabil Eng 1997;5:394–398. 5 Van Someren EJ, Vonk BF, Thijssen WA, et al: A new actigraph for long-term registration
100
6
7
8 9 10
of the duration and intensity of tremor and movement. IEEE Trans Biomed Eng 1998;45: 386–395. Bacher M, Scholz E, Diener HC: 24 hour continuous tremor quantification based on EMG recording. Electroencephalogr Clin Neurophysiol 1989;72:176–183. Carboncini MC, Manzoni D, Strambi S, et al: The relation between EMG activity and kinematic parameters strongly supports a role of the action tremor in parkinsonian bradykinesia. Mov Disord 2001;16:47–57. Spieker S, Boose A, Breit S, et al: Long-term measurement of tremor. Mov Disord 1998; 13(suppl 3):81–84. Burkhard PR, Langston JW, Tetrud JW: Voluntarily simulated tremor in normal subjects. Neurophysiol Clin 2002;32:119–126. Salarian A, Russmann H, Wider C, et al: Quantification of tremor and bradykinesia in
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
11 12
13
14
Parkinson’s disease using a novel ambulatory monitoring system. IEEE Trans Biomed Eng 2007;54:313–322. Elble RJ, Sinha R, Higgins C: Quantification of tremor with a digitizing tablet. J Neurosci Methods 1990;32:193–198. Goetz C, Poewe W, Rascol O, et al: The Unified Parkinson’s Disease Rating Scale (UPDRS): status and recommendations. Mov Disord 2003;18:738–750. Lukhanina EP, Kapoustina MT, Karaban IN: A quantitative surface electromyogram analysis for diagnosis and therapy control in Parkinson’s disease. Parkinsonism Relat Disord 2000;6:77–86. Geminiani G, Cesana BM, Tamma F, et al: Interobserver reliability between neurologists in training of Parkinson’s disease rating scales: a multicenter study. Mov Disord 1991; 6: 330– 335.
Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
References
Evaluation of Essential Tremor with Smartphones
21 Elble RJ, Pullman SL, Matsumoto JY, et al: Tremor amplitude is logarithmically related to 4- and 5-point tremor rating scales. Brain 2006;129:2660–2666. 22 Zhou XH, Obuchowski NA, McClish DK: Statistical Methods in Diagnostic Medicine. New York, Wiley, 2002. 23 Hanley JA, McNeil BJ: A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839–843. 24 Gescheider GA: Psychophysics: The Fundamentals. Mahwah, Erlbaum, 1997. 25 Elble RJ, Brilliant M, Leffler K, et al: Quantification of essential tremor in writing and drawing. Mov Disord 1996;11:70–78. 26 Matsumoto JY, Dodick DW, Stevens LN, et al: Three-dimensional measurement of essential tremor. Mov Disord 1999;14:288–294. 27 Mostile G, Giuffrida JP, Adam OR, et al: Correlation between Kinesia system assessments and clinical tremor scores in patients with essential tremor. Mov Disord 2010; 25: 1938– 1943.
28 Giuffrida JP, Riley DE, Maddux BN, et al: Clinically deployable Kinesia technology for automated tremor assessment. Mov Disord 2009;24:723–730. 29 LeMoyne R, Mastroianni T, Cozza M, et al: Implementation of an iPhone for characterizing Parkinson’s disease tremor through a wireless accelerometer application. Proc 32nd Int Conf of the IEEE Engineering in Medicine and Biology Society, Buenos Aires, 2010. 30 Kostikis N, Hristu-Varsakelis D, Arnaoutoglou M, et al: Towards remote evaluation of movement disorders via smartphones. Proc 33rd Int Conf of the IEEE Engineering in Medicine and Biology Society, Boston, 2011. 31 Joundi RA, Brittain JS, Jenkinson N, Green AL, Aziz T: Rapid tremor frequency assessment with the iPhone accelerometer. Parkinsonism Relat Disord 2011;17:288–290. 32 Daneault JF, Carignan B, Codère CÉ, Sadikot AF, Duval C: Using a smart phone as a standalone platform for detection and monitoring of pathological tremors. Front Hum Neurosci 2013;6:357.
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
101
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
15 Ginanneschi A, Degl’Innocenti F, Magnolfi S, et al: Evaluation of Parkinson’s disease: reliability of three rating scales. Neuroepidemiology 1988;7:38–41. 16 Richards M, Marder K, Cote L, et al: Interrater reliability of the Unified Parkinson’s Disease Rating Scale motor examination. Mov Disord 1994;9:89–91. 17 Worringham C, Rojek A, Stewart I: Development and feasibility of a smartphone, ECG and GPS based system for remotely monitoring exercise in cardiac rehabilitation. PLoS One 2011;6:e14669. 18 Dufau S, Dunabeitia JA, Moret-Tatay C, et al: Smart phone, smart science: how the use of smartphones can revolutionize research in cognitive science. PLoS One 2011; 6: e24974. 19 Senova S, Jarraya B, Iwamuro H, et al: Unilateral thalamic stimulation safely improved fragile X-associated tremor ataxia: a case report. Mov Disord 2012;27:797–799. 20 Garcin R: Contribution à l’étude sémiologique des tremblements intentionnels et des hyperkinésies volitionnelles. Rev Neurol (Paris) 1938;70:63–75.
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Using the Accelerometers Integrated in Smartphones to Evaluate Essential Tremor Suhan Senova a Damien Querlioz c Claire Thiriez b Pierre Jedynak a Béchir Jarraya a Stéphane Palfi a
a
Neurosurgery Department and b Neurology Department, Assistance Publique-Hôpitaux de Paris (APHP), Groupe Henri-Mondor Albert-Chenevier, Université Paris 12 UPEC, Faculté de Médecine, Créteil, and c Institut d’Electronique Fondamentale, Université Paris-Sud, CNRS, Orsay, France
Abstract Background/Aims: Evaluation of tremor constitutes a crucial step from the diagnosis to the initial treatment and follow-up of patients with essential tremor. The severity of tremor can be evaluated using clinical rating scales, accelerometry, or electrophysiology. Clinical scores are subjectively given, may be affected by intra- and interevaluator variations due to different experience, delays between consultations, and subtle changes in tremor severity. Existing medical devices are not routinely used: they are expensive, timeconsuming, not easily accessible. We aimed at showing that a smartphone application using the accelerometers embedded in smartphones is effective for quantifying the tremor of patients presenting with essential tremor. Methods: We developed a free iPhone/iPod application, Itremor, and evaluated different parameters on 8 patients receiving deep brain stimulation of the ventral intermediate nucleus of the thalamus: average and maximum accelerations, time above 1 g of acceleration, peak frequency, typical magnitude of tremor, for postural and action tremors, on and off stimulation. Results: We demonstrated good correlations between the parameters measured with Itremor and clinical score in all conditions. Itremor evaluation enabled higher discriminatory
© 2015 S. Karger AG, Basel 1011–6125/15/0932–0094$39.50/0 E-Mail [email protected] www.karger.com/sfn
power and degree of reproducibility than clinical scores. Conclusion: Itremor can be used for routine objective evaluation of essential tremor, and may facilitate adjustment of the treatment. © 2015 S. Karger AG, Basel
Introduction
Tremor constitutes a major motor symptom of several neurological diseases such as Parkinson’s disease and essential tremor. The severity of tremor can be evaluated using clinical rating scales, such as the Fahn-TolosaMarin score (Fahn score) [1]. Tremor can also be considered as a physical phenomenon, characterized through actigraphic, accelerometric, electromyographic, gyroscopic, or digitizing tablet recordings [2–11]. Precise evaluation of tremor is crucial for the diagnosis of neurological diseases, patient follow-up and the evaluation of treatment efficacy [12, 13]. The current clinical rating scales suffer from intra- and interrater variability that depends on several factors including rater experience and the delays between consultations. Furthermore, rating scales may not be sensitive enough to detect small variations of tremor [14–16]. Recordings made with current medical devices, while more objective, are not an optimal way to assess tremor either in clinical routine or for regular follow-up. Indeed, these devices are expensive, Suhan Senova and Pr. Stéphane Palfi Neurosurgery Department, Assistance Publique-Hôpitaux de Paris (APHP) Groupe Henri-Mondor Albert-Chenevier, Université Paris 12 UPEC Faculté de Médecine, FR–94010 Créteil (France) E-Mail stephane.palfi @ hmn.aphp.fr
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
Key Words Tremor · Essential tremor · Movement disorders · Accelerometry · Smartphone
from the three spectra. This analysis is only valid if the motion recorded is oscillatory enough. The application thus considered the spectrum analysis as valid only if the peaks were higher than an empirical threshold (equal to 15% of the spectral peak of patients with severe essential tremor scored 4/4 with Fahn score). When this validation criterion was met, the peak frequency parameter in Itremor was given by the average of its values in the three directions. Finally, the typical magnitude was computed in centimeters as mean acceleration divided by squared frequency and multiplied by a conversion factor taking into account scale changes and integration operation. The typical magnitude was meant as an estimation of the peak-to-peak movement amplitude of the hand in the tremor. Control experiments showed that this evaluation was effective for oscillatory movements. When the validation criterion was not met, the tremor magnitude was assigned the value 0.3 mm, which is the minimal distance of motion the human eye can see at a distance of 1 m. The five parameters, as well as the power spectra, were computed within this application Itremor and were accessible directly on the iPhone/iPod touch screen.
Apparatus We used an Apple iPod touch device. It is equipped with a three-axis accelerometer manufactured by ST Microelectronics as a standard feature. Our application performed recordings during 9-second periods of time with a sampling rate of 60 Hz. Within the application, it is possible to activate an ‘action tremor analysis’ mode, for which accelerometric data are high-pass filtered (one pole, 2.5-Hz cutoff) for the component of the signal due to the voluntary motion to be removed. When this mode was not activated, a 1-Hz cutoff was used to filter gravity. From the raw data recorded by the three-dimensional accelerometer, the application computed five parameters: the mean acceleration, the maximum acceleration, the time above 1 g, the peak frequency, and the typical magnitude of the tremor. The mean and maximum accelerations were computed directly from the filtered accelerometric data. The parameter time above 1 g is the time during each recording for which values of acceleration measured are greater than 9.8 m/s2, a standard accelerometric figure. The last two parameters were calculated by performing a spectral analysis. The spectrum density of the acceleration along the three x, y and z directions was computed with fast Fourier transform using Welch’s method with three overlapping segments and a Hamming window function. The highest peak was extracted
Clinical Assessment Eight patients, 3 women and 5 men, with an average age of 61 ± 13.5 years, were evaluated during a routine follow-up outpatient visit. This group of patients was homogeneously made of patients presenting with essential tremor. They all signed an informed consent for tremor assessment by Itremor and a videotaping for a blinded evaluation using the Fahn score. We received the agreement of the Ethics Committee of Henri Mondor University Hospital/Paris 12 Créteil, France. These patients suffered from essential tremor and were treated by uni- or bilateral Vim high-frequency electrical stimulation (table 1). Electrode implantation was previously described [19]. The iPod touch was attached to a dedicated armband tightly fixed around the wrist displaying the most severe tremor in the offstimulation condition (fig. 1). This armband is a commercial solution proposed by Apple in order to hold strongly the iPod while performing sport for instance. For objective evaluation of Itremor, during the consultation tremor was evaluated in on- and off-stimulation conditions while the patient was videotaped for subsequent blinded clinical score rating: thus, each patient was his own control. We evaluated postural tremor in the outstretched arms condition 3 times in a row, getting a triplet of values, and then, as a complement, in the swordsman condition also 3 times in a row. The swordsman condition consists in holding arms horizontally, with the elbows in semiflexion and supination, and with the fingers of each hand facing each other (fig. 1a) [20]: both distal and proximal postural tremors are involved. Results from outstretched arms and swordsman conditions were gathered under the name of postural tremor. Hand action tremor was evaluated 3 times in a row by asking the patients to go, pick up and give back a pen several times to the clinician in front of him. Between each evaluation, the patients could rest 1 min. Between the on- and off-stimulation series, the patients could rest 3 min. At the end, for subjective evaluation of Itremor, patients were asked the following questions: ‘Was the evaluation with Itremor tolerable?’ ‘Do you understand the use of Itremor?’ ‘Would you accept that your regular follow-up includes an evaluation with Itremor?’
Evaluation of Essential Tremor with Smartphones
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Methods
95
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
time-consuming and not easily accessible. The procedure may be poorly tolerated by the patient if performed on a routine basis, and the clinicians are required to learn elaborate connectors and software. As a consequence, specific recordings of tremor are now rarely resorted to, whereas they could be useful for both routine clinical follow-up and clinical research. For example, patients who are treated with deep brain stimulation of the ventral intermediate nucleus of the thalamus (Vim) for invalidating tremor need several consultations with a movement disorder specialist for adjusting to optimal stimulation parameters. Thus, there is a critical need for validating a medical device capable of evaluating tremor features objectively in clinical routine, being inexpensive, easily accessible, easy to use, and well tolerated by the patient. To address this clinical challenge, we developed a free iPhone/iPod application called Itremor that uses built-in accelerometers embedded in smartphones/MP3 players. We aimed at bringing the proof of concept that a smartphone application is effective for quantifying the tremor of patients diagnosed with essential tremor. We evaluated the clinical value of Itremor in such patients treated with electrical stimulation of the Vim. This is in line with the current idea of resorting to smartphones for medical and research purposes [17, 18].
b
Fig. 1. a Patient equipped with an armband containing the iPod and tightly fixed to her wrist. She is holding the swordsman position. b, c Printscreens of the iPod when the iPod application Itremor was used. b The tremor was recorded in the off-stimulation
c
Color version available online
a
condition and was scored 4/4 blindly by a neurologist. c The tremor was recorded in the on-stimulation condition and was scored 0/4 blindly by a neurologist. Each curve (red, green or blue) refers to one direction of space (colors refer to the online version only).
Table 1. Characteristics of the population of the study
Patient
Gender
Age, years
Date of start of stimulation
Recorded side
Parameters of monopolar stimulation: frequency (Hz)/pulse width (μs)/amplitude (V)/ impedance (Ω)/electrode site
1
female
60
2009
R
L: 130/60/2.5/491/1– R: 130/60/2.8/403/4–
2
male
58
1999
L
R: 185/90/3.2/127/4–
3
female
78
2008
L
L: 130/90/3.4/1,360/0– R: 130/90/2.5/491/4–
4
male
82
2006
R
L: 180/90/2.7/532/0– R: 180/90/2.3/450/5–
5
male
48
2010
L
L: 185/90/2.6/1,035/0– R: 185/90/2.3/913/4–
6
female
42
2010
R
L: 130/90/0.5/4,005/0– R: 130/90/1.8/711/4–
7
male
58
2006
L
L: 160/90/2.6/1,035/0– R: 160/90/2.8/267/4–
8
male
61
2010
L
L: 240/90/1.9/369/0– R: 240/90/1.9/751/4–
96
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
L = Left; R = right.
Color version available online
Itremor parameters
–4
n = 96 0
Action
log (Itremor parameter)
0
1 2 3 Clinical score (/4)
4
R = 0.83 p < 0.0001
–1
–2 n = 48 0
1
2
3
Clinical score (/4)
0 –1 –2 –3
n = 96 0
b
4
0
–1 n = 48 1
2
3
Clinical score (/4)
Fig. 2. Scatter plots (dots) and linear regression fit lines showing sig-
nificant correlation (R) between the Fahn clinical score and the natural logarithm of various Itremor parameters: mean acceleration (expressed in g), maximum acceleration (expressed in g) and tremor magnitude (expressed in cm), in two different conditions (postural,
Itremor data were stored at the end of each evaluation. Fahn scores were obtained blindly after randomization of the videos by a senior neurology specialist of movement disorders (P.J.). All videos were rated during the same session. Data Analysis Methods We first evaluated the clinical relevance of the parameters displayed by Itremor. We calculated the correlation between Fahn scores and the natural logarithm of mean acceleration, the natural logarithm of maximum acceleration, and the natural logarithm of typical magnitude of the tremor in the postural condition and in the action condition [21], pooling on- and off-stimulation condition measures from all the patients. Then, within each subgroup of measurements corresponding to each of the five grades of the Fahn scale, we calculated the percentage of measurements for which time above 1 g was superior to 1% of the time of recording, and for which a spectral analysis had been computed by Itremor. We resorted to Matlab software (MathWorks, Natick, Mass., USA) for this purpose. Finally, we compared the performance of the Fahn scores established by a movement disorder specialist expert in tremor and the Itremor measures as far as discriminatory power and degree of reproducibility (DR) of the measures are concerned. To this pur-
Evaluation of Essential Tremor with Smartphones
4
R = 0.71 p < 0.0001
1
0
e
1 2 3 Clinical score (/4)
R = 0.86 p < 0.0001
2 1 0 –1
n = 96 0
c
f
4
R = 0.83 p < 0.0001
1 0 –1
n = 48 0
4
1 2 3 Clinical score (/4)
1
2
3
4
Clinical score (/4)
action). For each linear regression, trials from all the patients and in on- and off-stimulation conditions have been pooled (n = 96 trials for the postural condition, n = 48 trials for the action condition). Correlation between Itremor parameters and clinical score was better for postural tremors (a–c) than for action tremor (d–f).
pose, for each condition and for each method of tremor evaluation, we computed areas under receiver operating characteristic (ROC) curves as a measure of discriminatory power [22] and performed statistical comparison between areas under paired ROC curves, resorting to MedCalc software (Mariakerke, Belgium) [23]. We defined DR as the ratio between the mean of the measure of all the trials and the mean of the triplet of the standard deviation of the measure. For a given condition, variation of DR was the ratio of the difference between the DR for a given parameter measured by Itremor and the DR of the Fahn score, divided by the DR of the Fahn score.
Results
In the postural condition, when pooling on- and offstimulation data from all the patients, a significant correlation was found (fig. 2) between Fahn scores and the natural logarithm of mean acceleration (r = 0.92, p < 10–4, n = 96), Fahn scores and the natural logarithm of maximum acceleration (r = 0.89, p < 10–4, n = 96), Fahn scores Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
97
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
–3
R = 0.89 p < 0.0001
log (Itremor parameter)
–2
a
d
log (Itremor parameter)
–1
log (Itremor parameter)
Postural
log (Itremor parameter)
0
Tremor magnitude
Maximum acceleration
1
R = 0.92 p < 0.0001
log (Itremor parameter)
Mean acceleration
1
Action
80 Sensitivity
40
0
0 0
60
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
0
100 – Specificity
and the natural logarithm of tremor magnitude (r = 0.86, p < 10–4, n = 96). Similarly, in the action condition, a significant correlation was found (fig. 2) between Fahn scores and the natural logarithm of mean acceleration (r = 0.83, p < 10–4, n = 48), Fahn scores and the natural logarithm of maximum acceleration (r = 0.71, p < 10–4, n = 48), Fahn scores and the natural logarithm of tremor magnitude (r = 0.83, p < 10–4, n = 48). For all the clinical situations rated 0, 1 or 2/4 on the Fahn tremor rating scale, the time above 1 g was never more than 1% of the recording time. This parameter was relevant to assess tremors rated 3 or 4/4 (table 2). When a tremor was scored 0 or 1/4 on the Fahn scale, it never passed the validation criterion which allows spectral analysis to be computed (table 2). Discriminatory powers between on- and off-stimulation condition recordings were significantly higher (p < 0.05) for the parameters mean acceleration and maximum acceleration from Itremor than for the Fahn score in the postural condition (fig. 3). We observed that the DR of the mean acceleration and the tremor magnitude were improved in all conditions compared with the DR of the Fahn score (table 3). The DR of the maximum acceleration was lowered compared with the DR of the Fahn score for the action condition, and was improved for the postural condition. A hundred percent of patients found that the evaluation with Itremor was tolerable, understood the use of Itremor and accepted a regular evaluation with Itremor during follow-up. 98
40
60 100 – Specificity
Clinical score log (mean acceleration)
log (maximum acceleration) log (tremor magnitude)
Table 2. Relevance of the Itremor parameter time above 1 g and the spectral analysis for each grade of the Fahn scale in two conditions (postural, action) Clinical score
0 (n = 26 trials) 1 (n = 56 trials) 2 (n = 32 trials) 3 (n = 15 trials) 4 (n = 15 trials)
Trials for which time above Trials for which spectral 1 g >1% of recorded time, % analysis was performed, % postural
action
postural
action
0 0 0 10 100
0 0 0 60 100
0 0 78 100 100
0 0 46 100 100
Discussion
All the parameters given by Itremor were clinically correlated for both the postural and action components of tremor in patients suffering from essential tremor. The use of Itremor for tremor assessment was well tolerated and well accepted by patients. To define the parameters to be determined by Itremor, we were guided by the functional impact of tremor, looking at patients drinking a glass of water. The difficulties encountered are mostly due to the acceleration to which the water is submitted inside the glass without spilling over and to the magnitude of the oscillations of the glass when approaching the mouth. Thus, we calculated parameters directly related to the acceleration such as the mean acceleration, the maximum acceleration and the time above 1 g, and also parameters coming from the Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
tremor assessment, for postural and action tremor separately. Curves for which the area under the ROC curve is statistically (p < 0.05) different from the area under the ROC curve for clinical score rating are indicated with an asterisk. Discriminatory powers between on- and off-stimulation condition recordings were significantly higher (p < 0.05) for log (mean acceleration) and log (maximum acceleration) from Itremor than for the Fahn score in the postural condition.
Sensitivity
80
Fig. 3. ROC curves for each modality of
Color version available online
Postural
spectral analysis such as the peak frequency and the magnitude of the tremor. The armband to which the iPod was attached was fixed around the wrist, but could also be attached around the hand or the elbow depending on clinical specificities of the patient’s tremor. Mean and maximum accelerations are parameters that are simple to compute and to understand. But maximum acceleration may be biased by short periods of violent accelerations or by maximum acceleration exceeding the capacity of the accelerometer of the iPod touch. The parameter time above 1 g should enable us to better characterize and follow up patients with violent tremors; table 2 shows that this parameter is specifically relevant for tremors rated 3 or 4 on the Fahn scale. In clinical practice, treatment efficacy measured on a violent tremor may be better estimated by the reduction of periods of high acceleration. Three other parameters were computed after spectral analysis. For this kind of analysis to be physically valid, it has to be performed only in clinical situations where the motion of the wrist could be estimated as oscillatory, even if only transiently, by a clinician. From this idea, we elaborated a criterion on the value of the peak of the spectral analysis for the spectral analysis to be regarded as valid, then peak frequency and tremor magnitude to be calculated and displayed on the screen of the iPod. The validity of this criterion was confirmed by the results in table 2: no motion rated 0 or 1 on the Fahn scale justified spectral analysis, all motions rated 3 and 4 justified spectral analysis. In this study, we quantitatively compared the performance of Itremor with a validated clinical scale (Fahn score). We looked for a logarithmic correlation between mean acceleration, maximum acceleration, tremor magnitude calculated by Itremor and clinical scores, as observed in previously published studies with other tools of
tremor assessment [21, 24–28]. For all these parameters and in all the conditions, a significant correlation with the Fahn score was found. Testing patients in on- and offstimulation conditions enabled us to evaluate the correlation between Itremor parameters and clinical score, for tremors clinically scored from 0 to 4, as detailed in table 2. In the next study, we will evaluate whether using Itremor for the adjustment of deep brain stimulation parameters, more subtle than on- versus off-stimulation parameter adjustments, enables better outcomes of Vim deep brain stimulation. The study also demonstrates that a medical device with an accelerometer and appropriate data processing offers higher discriminatory power for postural tremor. We describe improved evaluation reproducibility of the tremor of a given patient by a given evaluator performing multiple evaluations at a given time. These findings were imaginable due to the finer graduations of an accelerometric data scale with a sensitivity of 20 mg of acceleration compared with a clinical scale with only five levels. Characterizing action tremor required not spoiling the analysis with the voluntary component of movement. Fortunately, this component was typically in a frequency range inferior to 2.5 Hz for the task we asked the patients to perform: so signal was high-pass filtered before the parameters were computed. The correlation coefficients between the various parameters from Itremor and the clinical scores were thus good even if slightly lower for action than for postural tremor: the frequential impoverishment of the signal after high-pass filtering might be one of the causes. However, the difficulty to score that type of tremor clinically may be another cause: during the movement, tremor may vary, and it may also be difficult for both the patient and the clinician to reproducibly perform the task. As far as discriminatory power is concerned, a superiority of Itremor parameters over clinical score was not found for action tremor: not only the lower correlation between Itremor parameters and clinical score for action tremor may account for it, but also comparing two sets of stimulation parameters and their clinical effect with less difference than on- versus off-stimulation parameters may have helped revealing a higher discriminatory power for Itremor parameters than for clinical score parameters when evaluating action tremor. Other groups tried to develop similar iPhone applications to quantify tremor [29, 30]. However, in contrast to the Itremor application, these applications were not yet validated clinically. In addition, our application included modalities for assessing not only postural trem-
Evaluation of Essential Tremor with Smartphones
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
celeration, maximum acceleration and tremor magnitude, relative to clinical Fahn score in two conditions (postural, action) Condition
Itremor parameter
Variation of DR, %
Postural (n = 96 trials)
mean acceleration maximum acceleration tremor magnitude
61.7 16.8 29.2
Action (n = 48 trials)
mean acceleration maximum acceleration tremor magnitude
61.6 –429.1 2.5
Data from all the patients in on- and off-stimulation conditions have been pooled.
99
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
Table 3. Variation of DR for three Itremor parameters: mean ac-
or as in the work of LeMoyne et al. [29] and Kostikis et al. [30], but also action tremors. Moreover, Itremor has the advantage of being completely autonomous and the whole signal analysis is performed and displayed on the iPhone/iPod directly, without requiring any data transfer to an external computer by e-mail [29], or being Web-based and running on a distant computer [30]. As a result, Itremor is easily usable in the context of outpatient clinics or hospitalization by clinicians and nurses. The iPhone application called ‘iSeismo’ was also used to quantify tremor [31]. However, this application was originally developed for earthquake detection and not for clinical use, whereas Itremor’s parameters aimed at quantifying specific clinically relevant motions, with algorithms adjusted to take into account clinical orders of magnitude. The first study showing a clinical validation of a smartphone application was published in 2013 [32]. In this study, the authors set up their clinical rating scale and did not refer to an internationally recognized clinical tremor rating scale. In addition, the author recognized their inability to detect short moments of intense and incapacitating tremor which were easily quantified by Itremor using the parameter called ‘time above 1 g’. Several other iPhone applications which were not clinically validated nor described in any scientific paper can be found on the Apple Store [Itrem (2011), ParkinsonMeter (2012), LiftPulse (2013) and StudyMyTremor (2014)]. Itremor does not compete with electromyography for precise diagnosis. They rather complement each other: Itremor may facilitate the routine objective eval-
uation of essential tremor and facilitate adjustment of the treatment. Itremor is a free application, distributed on request according to the Ad Hoc process defined by Apple. Smartphones are widely available and used by medical doctors and nurses who could have their own smartphone equipped with Itremor, ready to be used extensively and easily in the outpatient clinic and the inpatient department and even during clinical trials. Itremor is easy to use, and a detailed processing of data is done and displayed directly on the screen of the iPhone without the need for expert computer skills or signal processing. Data are also easy to save and export to a computer with secured transfer methods. Itremor highlights the potential impact of smartphones for modern, inexpensive, and efficient healthcare.
Conclusion
Evaluating essential tremor with the iPhone application Itremor was validated. It may improve tremor evaluation, follow-up, and treatment adjustment and facilitate clinical research. Itremor and electromyography complement each other and do not compete. Acknowledgment We thank the Association pour la Recherche sur la Stimulation Cérébrale (ARSC), Dr. Liza Leventhal, Dr. Olivier Bill and Arthur Silve for reviewing the manuscript.
1 Fahn S, Tolosa E, Marin C: Clinical rating scale for tremor; in Jankovic J, Tolosa E (eds): Parkinson’s Disease and Movement Disorders. Baltimore, Williams & Wilkins, 1993, pp 225–234. 2 Gerr F, Letz R, Green RC, et al: Relationships between quantitative measures and neurologist’s clinical rating of tremor and standing steadiness in two epidemiological studies. Neurotoxicology 2000;21:753–760. 3 Hoff JI, Wagemans EA, van Hilten BJ: Ambulatory objective assessment of tremor in Parkinson’s disease. Clin Neuropharmacol 2001; 24:280–283. 4 Van Someren EJ: Actigraphic monitoring of movement and rest activity rhythms in aging, Alzheimer’s disease, and Parkinson’s disease. IEEE Trans Rehabil Eng 1997;5:394–398. 5 Van Someren EJ, Vonk BF, Thijssen WA, et al: A new actigraph for long-term registration
100
6
7
8 9 10
of the duration and intensity of tremor and movement. IEEE Trans Biomed Eng 1998;45: 386–395. Bacher M, Scholz E, Diener HC: 24 hour continuous tremor quantification based on EMG recording. Electroencephalogr Clin Neurophysiol 1989;72:176–183. Carboncini MC, Manzoni D, Strambi S, et al: The relation between EMG activity and kinematic parameters strongly supports a role of the action tremor in parkinsonian bradykinesia. Mov Disord 2001;16:47–57. Spieker S, Boose A, Breit S, et al: Long-term measurement of tremor. Mov Disord 1998; 13(suppl 3):81–84. Burkhard PR, Langston JW, Tetrud JW: Voluntarily simulated tremor in normal subjects. Neurophysiol Clin 2002;32:119–126. Salarian A, Russmann H, Wider C, et al: Quantification of tremor and bradykinesia in
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
11 12
13
14
Parkinson’s disease using a novel ambulatory monitoring system. IEEE Trans Biomed Eng 2007;54:313–322. Elble RJ, Sinha R, Higgins C: Quantification of tremor with a digitizing tablet. J Neurosci Methods 1990;32:193–198. Goetz C, Poewe W, Rascol O, et al: The Unified Parkinson’s Disease Rating Scale (UPDRS): status and recommendations. Mov Disord 2003;18:738–750. Lukhanina EP, Kapoustina MT, Karaban IN: A quantitative surface electromyogram analysis for diagnosis and therapy control in Parkinson’s disease. Parkinsonism Relat Disord 2000;6:77–86. Geminiani G, Cesana BM, Tamma F, et al: Interobserver reliability between neurologists in training of Parkinson’s disease rating scales: a multicenter study. Mov Disord 1991; 6: 330– 335.
Senova/Querlioz/Thiriez/Jedynak/Jarraya/ Palfi
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
References
Evaluation of Essential Tremor with Smartphones
21 Elble RJ, Pullman SL, Matsumoto JY, et al: Tremor amplitude is logarithmically related to 4- and 5-point tremor rating scales. Brain 2006;129:2660–2666. 22 Zhou XH, Obuchowski NA, McClish DK: Statistical Methods in Diagnostic Medicine. New York, Wiley, 2002. 23 Hanley JA, McNeil BJ: A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839–843. 24 Gescheider GA: Psychophysics: The Fundamentals. Mahwah, Erlbaum, 1997. 25 Elble RJ, Brilliant M, Leffler K, et al: Quantification of essential tremor in writing and drawing. Mov Disord 1996;11:70–78. 26 Matsumoto JY, Dodick DW, Stevens LN, et al: Three-dimensional measurement of essential tremor. Mov Disord 1999;14:288–294. 27 Mostile G, Giuffrida JP, Adam OR, et al: Correlation between Kinesia system assessments and clinical tremor scores in patients with essential tremor. Mov Disord 2010; 25: 1938– 1943.
28 Giuffrida JP, Riley DE, Maddux BN, et al: Clinically deployable Kinesia technology for automated tremor assessment. Mov Disord 2009;24:723–730. 29 LeMoyne R, Mastroianni T, Cozza M, et al: Implementation of an iPhone for characterizing Parkinson’s disease tremor through a wireless accelerometer application. Proc 32nd Int Conf of the IEEE Engineering in Medicine and Biology Society, Buenos Aires, 2010. 30 Kostikis N, Hristu-Varsakelis D, Arnaoutoglou M, et al: Towards remote evaluation of movement disorders via smartphones. Proc 33rd Int Conf of the IEEE Engineering in Medicine and Biology Society, Boston, 2011. 31 Joundi RA, Brittain JS, Jenkinson N, Green AL, Aziz T: Rapid tremor frequency assessment with the iPhone accelerometer. Parkinsonism Relat Disord 2011;17:288–290. 32 Daneault JF, Carignan B, Codère CÉ, Sadikot AF, Duval C: Using a smart phone as a standalone platform for detection and monitoring of pathological tremors. Front Hum Neurosci 2013;6:357.
Stereotact Funct Neurosurg 2015;93:94–101 DOI: 10.1159/000369354
101
Downloaded by: NYU Medical Center Library 128.122.253.212 - 4/20/2015 6:54:38 PM
15 Ginanneschi A, Degl’Innocenti F, Magnolfi S, et al: Evaluation of Parkinson’s disease: reliability of three rating scales. Neuroepidemiology 1988;7:38–41. 16 Richards M, Marder K, Cote L, et al: Interrater reliability of the Unified Parkinson’s Disease Rating Scale motor examination. Mov Disord 1994;9:89–91. 17 Worringham C, Rojek A, Stewart I: Development and feasibility of a smartphone, ECG and GPS based system for remotely monitoring exercise in cardiac rehabilitation. PLoS One 2011;6:e14669. 18 Dufau S, Dunabeitia JA, Moret-Tatay C, et al: Smart phone, smart science: how the use of smartphones can revolutionize research in cognitive science. PLoS One 2011; 6: e24974. 19 Senova S, Jarraya B, Iwamuro H, et al: Unilateral thalamic stimulation safely improved fragile X-associated tremor ataxia: a case report. Mov Disord 2012;27:797–799. 20 Garcin R: Contribution à l’étude sémiologique des tremblements intentionnels et des hyperkinésies volitionnelles. Rev Neurol (Paris) 1938;70:63–75.
